The human brain comprises about a hundred billion neurons connected through quadrillion synapses. Spiking neural networks (SNNs) take inspiration from the brain to model complex cognitive and learning tasks. Neuromorphic engineering implements SNNs in hardware, aspiring to mimic the brain at scale (i.e., 100 billion neurons) with biological area and energy efficiency. The design of ultra-energy-efficient and compact neurons is essential for the large-scale implementation of SNNs in hardware. In this article, we have experimentally demonstrated a partially depleted (PD) silicon-on-insulator (SOI) MOSFET-based leaky integrate-and-fire (LIF) neuron, where energy efficiency and area efficiency are enabled by two elements of the design–first is the tunneling-based operation and the second is a compact subthreshold SOI control circuit design. Band-to-band tunneling (BTBT)-induced hole storage in the body is used for the “integrate” function of the neuron. A compact control circuit “fires” a spike when the body’s potential exceeds the firing threshold. The neuron then “resets” by removing the stored holes from the body contact of the device. Additionally, the control circuit provides “leakiness” in the neuron, which is an essential property of the biological neurons. The proposed neuron provides $10\times $ higher area efficiency compared to the CMOS design with equivalent energy/spike. Alternatively, it has a $10^{4}\times $ higher energy efficiency at area-equivalent neuron technologies. Biologically comparable energy efficiency and area efficiency, along with CMOS compatibility, make the proposed device attractive for large-scale hardware implementation of SNNs.

中文翻译：

---

基于带间隧道的超节能硅神经元

人类大脑由大约一千亿个神经元组成，这些神经元通过亿万个突触相连。尖峰神经网络 (SNN) 从大脑中汲取灵感来模拟复杂的认知和学习任务。神经拟态工程在硬件中实现 SNN，希望以生物面积和能源效率大规模模拟大脑（即 1000 亿个神经元）。超节能和紧凑型神经元的设计对于在硬件中大规模实现 SNN 至关重要。在本文中，我们通过实验展示了一种基于部分耗尽 (PD) 绝缘体上硅 (SOI) MOSFET 的泄漏集成和激发 (LIF) 神经元，其中能量效率和面积效率由设计的两个元素实现——首先是基于隧道的操作，其次是紧凑的亚阈值 SOI 控制电路设计。体内带间隧道 (BTBT) 诱导的空穴存储用于神经元的“整合”功能。当人体电位超过触发阈值时，紧凑的控制电路会“触发”尖峰信号。然后神经元通过从设备的身体接触中移除存储的孔来“重置”。此外，控制电路在神经元中提供“泄漏”，这是生物神经元的基本特性。与具有等效能量/尖峰的 CMOS 设计相比，所提出的神经元提供了 10 美元/倍的面积效率。或者，在面积等效的神经元技术中，它的能效提高了 10 美元 ^{4}\x 美元。生物可比的能源效率和面积效率，以及 CMOS 兼容性，使所提出的设备对 SNN 的大规模硬件实现具有吸引力。